Google formally killed Federated Learning of Cohorts last week. The replacement is Topics API, announced January 25, and if you squint at it hard enough, it looks like a reasonable compromise between user privacy and advertiser utility. Don’t squint. Look at it straight on, because what Topics API actually represents is a dramatic compression of the targeting universe that programmatic buyers have spent fifteen years building on.

FLoC, for those who have been tracking this saga since 2019, was Google’s first stab at a cookieless interest-targeting mechanism. It would cluster users into cohorts based on their browsing history, with each cohort assigned a numeric ID. The privacy community hated it — the Electronic Frontier Foundation called it a “privacy nightmare” almost immediately — and W3C objections piled up fast. By March 2021, the criticism had reached a point where even Google’s own browser team couldn’t defend it publicly. FLoC is dead. Topics API is what comes next.

What Topics API Actually Is

The mechanism is simpler than FLoC, and that simplicity is both the point and the problem. Chrome observes a user’s browsing activity over a rolling three-week window, then assigns that user to up to three “topics” drawn from a taxonomy of 350 categories. When a user visits a publisher site that participates in the Privacy Sandbox, the browser can share one topic from each of the past three weeks with the page’s ad tech partners.

Three hundred fifty categories. To put that in context: LiveRamp’s data marketplace has thousands of audience segments available. Nielsen’s syndicated taxonomy runs into the tens of thousands. Even the IAB’s standard content taxonomy, which is designed for contextual classification, not behavioral targeting, has around 700 tier-two categories. Google is proposing to replace the behavioral signal infrastructure that underpins a multi-hundred-billion-dollar industry with a 350-item lookup table.

The categories themselves are broad. You get “Autos & Vehicles,” not “SUV shoppers in-market within 60 days.” You get “Finance/Banking,” not “high-net-worth households researching wealth management.” You get “Sports/Basketball,” not “NBA playoff viewers aged 25-44 who have visited Nike.com in the past 30 days.” For brand awareness at scale, Topics might be serviceable. For the performance campaigns that actually drive platform revenue — the retargeting, the lower-funnel push, the competitive conquesting — it is not a replacement. It is a lesser substitute.

The Reach-Relevance Trade-Off Is Getting Worse

There is a version of this argument where you accept the targeting compression and compensate with reach. If you can’t find the exact audience you want, you buy a larger, blunter audience and rely on creative, landing page optimization, and offer quality to do the qualification work. Some advertisers have always operated this way. Direct response brands that live in MER math rather than audience sophistication are less dependent on behavioral depth.

But that version of the trade-off gets harder under Topics for a structural reason: you don’t control the audience. With third-party cookies, even with all their known flaws — ID fragmentation, cross-device gaps, bot pollution — you could build proprietary segments, layer deterministic first-party data, and construct audience models with genuine differentiation. Your data strategy was defensible competitive advantage.

With Topics API, every advertiser buying against “Autos & Vehicles” is targeting the same signal. There is no segment building, no third-party enrichment, no look-alike extension. Google’s Privacy Sandbox explainer for Topics makes clear that the API is deliberately designed to limit signal specificity. That is a feature for privacy advocates. For advertisers competing in the same category, it is a feature that eliminates audience differentiation entirely.

How This Changes the Walled Garden Dynamic

Google will tell you that first-party data within Google properties — Search intent, YouTube watch behavior, Gmail inference — is fully available and unaffected by Topics API. That is true. It is also the most convenient possible outcome for Google’s own advertising business. The mechanism that limits third-party audience intelligence happens to push advertisers toward Google’s own data assets, which are exempt from the Privacy Sandbox restrictions.

This is not a conspiracy theory. It is the logical structural consequence of a privacy framework designed by the company that also operates the dominant search, video, and display advertising platforms. The UK Competition and Markets Authority noted exactly this concern in its 2021 investigation, which led to the commitments Google made to avoid preferential treatment for its own ad business. Whether those commitments are enforceable in practice is a separate question, and one that regulators are still working through.

For programmatic buyers, the practical effect is real: targeting precision on open-web inventory will decline significantly under Topics API, while targeting precision on Google-owned properties stays constant or improves. Budget allocation will rationally shift toward Google properties and other walled gardens with equivalent first-party data depth. The open web gets commoditized further.

What Advertisers Should Actually Be Doing Now

The honest answer is that Topics API is not the problem you should be solving for in January 2022. Cookie deprecation is still targeted for late 2023 at earliest. The problem you should be solving for is the gap between your current audience targeting infrastructure and whatever post-cookie world actually materializes — because Topics is one scenario, not the only one.

First-party data investment remains the highest-leverage action regardless of which Privacy Sandbox API wins. If you can build authenticated audiences from your own customer data — email captures, login walls, CRM matching — you are less dependent on browser-level signals whether those signals are cookies or Topics. Publishers who have invested in registration walls and identity solutions are better positioned than those who have not, independent of Google’s timeline.

Second, clean room infrastructure is worth building now. Amazon Marketing Cloud, LiveRamp’s Safe Haven, and emerging clean room platforms from InfoSum and Habu enable audience-level analysis without individual-level ID porting. This is where the sophisticated measurement work will happen in a post-cookie environment.

Third, contextual targeting deserves genuine reinvestment. Contextual has historically been treated as a lower-cost alternative to behavioral, and the gap in performance was real when behavioral signals were abundant. That gap narrows when behavioral signals compress to 350 categories. Companies like Seedtag and Peer39 have built contextual intelligence infrastructure that goes significantly beyond category matching, incorporating semantic analysis and real-time page scoring. This is worth evaluating seriously, not just as a backup plan.

The announcement of Topics API is not the final word on how this plays out. The W3C process continues, Privacy Sandbox APIs will change, and there is a realistic scenario where cookie deprecation slips again. But treating that as permission to delay first-party data investment would be the mistake. Topics API tells you something real about Google’s direction, even if the specific implementation shifts.

FAQ

What is the difference between FLoC and Topics API? FLoC assigned users to cohort IDs based on browsing history clusters, then shared those IDs with ad tech partners. Topics API assigns users to broad interest categories from a fixed 350-item taxonomy and shares up to three topics per page load. Topics is more transparent and less fingerprint-able than FLoC, but also delivers significantly less targeting specificity.

Does Topics API replace third-party cookies completely? No. Topics API handles interest-based targeting, but the Privacy Sandbox also includes separate proposals for attribution (Attribution Reporting API) and fraud prevention (Trust Tokens). Third-party cookies handle multiple functions, and there is no single replacement — rather a collection of APIs each addressing a different use case.

When will Topics API be fully deployed in Chrome? Google has not given a specific deployment timeline for Topics API beyond “part of Privacy Sandbox.” Cookie deprecation is currently targeted for late 2023, but that timeline has already shifted once and may shift again.

Should advertisers start testing Topics API now? Testing infrastructure is available in Chrome developer flags, but the API is not in production traffic yet. The more urgent investment is in first-party data and clean room infrastructure that will provide value regardless of which API configuration ultimately deploys.